The DOI matches its Crossref record; 17 of the 17 references it cites were found in a registry or an index.
ProcgenOOD: a benchmark for out-of-distribution generalization in reinforcement learning
Abstract
The Procgen benchmark is the primary testbed for studying reinforcement learning (RL) generalization, but it is limited to in-distribution (ID) evaluation, where train and test levels are sampled from the same distribution with no mechanism to isolate or control specific environment features. We introduce ProcgenOOD, an extension of Procgen that enables principled out-of-distribution (OOD) evaluation through configurable holdout variables.* ProcgenOOD creates true train/test splits over background images, platform difficulties, wall textures, and agent and enemy sprites, with adjustable holdout fractions and interpolation or extrapolation sampling modes. We establish baselines using PPO with NatureCNN and ImpalaCNN architectures, data augmentation, and exploration methods previously evaluated on Procgen. Our experiments reveal that ID and OOD rankings diverge: configurations that excel under standard Procgen evaluation degrade under OOD conditions, underscoring the need for benchmarks with explicit distributional control. *Code: https://github.com/USArmyResearchLab/ProcgenOOD
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 1
- Raffin, 2021, Stable-baselines3: Reliable reinforcement learning implementations, Journal of Machine Learning Research, 22, 1
Metadata
- Authors
- Corder, K., Park, S. J., & Vindiola, M. M.
- Affiliations
- Parsons (United States) (US)
- DEVCOM Army Research Laboratory (US)
- ORCID
- Vindiola, Manuel M.
- Venue
- Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications VIII
- Publisher
- SPIE
- Published
- 2026-06-11
- Pages
- 13
- OpenAlex
- W7164302410
- Semantic Scholar
- d147f31a15d5b8eba1fab0a2d14a2b75e2b7c4a1
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
- Keywords
- out-of-distribution generalization, benchmark design, interpolation and extrapolation, PPO, data augmentation, exploration methods, video game environments
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: ProcgenOOD: a benchmark for out-of-distribution generalization in reinforcement learning. https://app.cochat.ai/papers/10.1117/12.3093362
- The paper
- Corder, K., Park, S. J., & Vindiola, M. M. (2026). ProcgenOOD: a benchmark for out-of-distribution generalization in reinforcement learning. Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications VIII. https://doi.org/10.1117/12.3093362
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.